Auto-Annotation with SAM and Custom Models

Hi everyone, I write because I read the tutorial at this link: https://docs.cvat.ai/docs/manual/advanced/serverless-tutorial/#debugging-a-serverless-function and also searched through the repo but I can't find the information I'm looking for.

Since the documentation says that you can add your own custom models as well, I would like to be able to use my own segmentation network written by me.

The problem is that I don't understand how to write the function.yaml file, I've tried looking at the segmentation templats in the /serverless folder I can't figure out how the models get input from CVAT.

For example, for SAM, which code is here https://github.com/cvat-ai/cvat/tree/develop/serverless/pytorch/facebookresearch/sam/nuclio where the model uses the user's prompt? What can I do should I want to use a similar approach and give my network a selected area of the whole image?